How to Stop Worrying the Experimental
Uncertainty of Heterogeneous Bioactivity Data
This research delves into the critical role of uncertainty quantification in autonomous decision-making, particularly in integrating machine learning with chemistry automation for an autonomous design-make-test-analyze cycle. Despite the focus on prediction uncertainty in existing literature, this study emphasizes the uncertainty of experimental data, using the ChEMBL database as a primary data source. We highlight the significant variability in bioactivity measurements due to differences in experimental design, methods, and potential data annotation errors, which impact the accuracy of in silico models.
Our analysis centers on cross-platform and within-site variability of experimental measurements, employing a data processing protocol to ensure high-confidence equilibrium constants (Ki) and half-maximal inhibitory concentration (IC50) values. The study assesses individual agreement between measurements using log-potency difference and explores the effects of assay platform selection on experimental uncertainty. Our findings reveal significant differences in uncertainty between cross-platform measurements and those made under the same assay platform, as well as the impact of assay setup modifications on measurement reliability.
By applying rigorous data processing and uncertainty evaluation methods, including statistical analyses in Python and R, this research proposes a nuanced understanding of experimental uncertainty in bioactivity data. The conclusions drawn from this study not only advance the field of uncertainty quantification in chemical and biological research but also offer a framework for improving the design and analysis of future experiments.
Research Area | Presenter | Title | Keywords |
---|---|---|---|
Probability, Statistics, and Machine Learning | Waghe, Shreyas | Fair Machine Learning (0.827586), Machine Learning (0.923077) | |
Computer Science | Berduo, Alan Jesse | Machine learning | |
Engineering | Li, Agnes | Machine learning | |
Computer Science | Shaikh, Aymaan | Machine learning | |
Mathematics and Statistics | Burns, Benjamin | machine learning |